An Accurate Prediction of Coronary Heart Disease Using Ensemble Algorithms

Abstract

Coronary Heart-disease (CHD) is one the foremost causes of death globally, making the accurate recognition of it vital. Machine-learning (ML) and deep-learning (DL) are two of the newest technologies being put to use in the healthcare industry. This paper uses a risk factor method to examine how CHD can be predicted. K-Nearest-Neighbors, Binary-Logistic Classification, and Naïve-Bayes are only few of the predictive methods that are measured by accuracy, recall, and receiver operating characteristic (ROC) curves for evaluation. Ensemble modeling strategies including bagging, boosting, and stacking are contrasted to these foundational classifiers. The efficacy of using ensemble approaches to enhance coronary heart disease prediction was investigated using a comparative analytical methodology. The ‘Cardiovascular-Disease Dataset,’ which includes 80,000 records of patient data for CHD, is used to evaluate the modeling techniques. It is demonstrated that bagged models, on average, are more accurate than their non-bagged counterparts by the margin of 1.99 percentage points. The highest accuracy of all models was 76.5%, but the AUC for the boosted models was 0.78. Accuracy of 75.1% was achieved by the stacked model consisting of KNN, random-forest classifier, and SVM. In addition, data-analytic methods and K-Folds cross-validation were used to verify the tested models' performance.

Topics & Concepts